UC San Diego

In this dissertation, we investigate problems of resource allocation for uplink multiuser wireless communication channels. The users want to upload compressed video streams with different rate-distortion (RD) characteristics to a base station via a multi-carrier channel. The video is organized in groups of pictures (GOP), and the base station makes resource allocation decisions based jointly on the channel state information (CSI) and the RD information to minimize the average video distortion across all users. We first study the case where the channel is slowly varying. With the assumption of the channel staying constant over the duration of a GOP, we derive an optimal condition for minimizing the sum of distortions in a setting of continuous frequency channel response. Our result shows that for a specific band to be assigned to a particular user, the product of the slope of the RD curve of that user and a marginal rate increment corresponding to that band should be maximized. We then design an algorithm which iteratively calculates the status of the RD information and the CSI in a block fading setting. Compared to two baseline resource allocation algorithms that use only a single layer of information, the simulation results show that the cross layer scheme can consistently improve the performance of the system. We then focus on the case of a channel with arbitrary mobility. To better utilize the time varying channel, we design a pilot symbol assisted modulation (PSAM) system which periodically adapts the modulation format according to the instantaneous CSI. We show that the two baseline algorithms with single layer information behave quite differently as the Doppler spread changes, but the cross layer algorithm performs robustly across all different mobile speeds. We further characterize the capacity gain of using the cross layer scheme